<p>Using the wisdom of crowds---combining many individual forecasts to obtain an aggregate estimate---can be an effective technique for improving forecast accuracy. When individual forecasts are drawn from independent and identical information sources, a simple average provides the optimal crowd forecast. However, correlated forecast errors greatly limit the ability of the wisdom of crowds to recover the truth. In practice, this dependence often emerges because information is shared: forecasters may to a large extent draw on the same data when formulating their responses. </p><p>To address this problem, I propose an elicitation procedure in which each respondent is asked to provide both their own best forecast and a guess of the average forecast that will be given by all other respondents. I study optimal responses in a stylized information setting and develop an aggregation method, called pivoting, which separates individual forecasts into shared and private information and then recombines these results in the optimal manner. I develop a tailored pivoting procedure for each of three information models, and introduce a simple and robust variant that outperforms the simple average across a variety of settings.</p><p>In three experiments, I investigate the method and the accuracy of the crowd forecasts. In the first study, I vary the shared and private information in a controlled environment, while the latter two studies examine forecasts in real-world contexts. Overall, the data suggest that a simple minimal pivoting procedure provides an effective aggregation technique that can significantly outperform the crowd average.</p> / Dissertation

Tidigare forskning visar på att analytiker i många fall begår fel vid framtagandet av estimat, samtidigt som deras estimat anses vara en proxy för vad marknaden förväntar sig. Denna studie avser att undersöka om analytiker är den främsta representanten för marknaden eller om Crowdsourcing, som en tillämpning av Wisdom of Crowds, kan representera marknaden bättre. Jämförelsen görs genom en multipel regression där 117 observationer används för att undersöka om konsensusestimat från crowdsourcing-tjänsten Pinpoint Estimates åstadkommer en större kursreaktion än analytikers konsensusestimat, givet ett visst oväntat resultat. Jämförelser görs även för träffsäkerhet där vi genom 123 observationer undersöker vilken grupps konsensusestimat som är mer träffsäker kontra utfallen. Studien finner indikationer på att Pinpointestimat representerar marknaden bättre samt att Pinpoint generellt är mer träffsäkra än analytiker.

Enabled by the wave of online crowdsourcing activities, extracting the Wisdom of Crowds (WoC) has become an emerging research area, one that is used to aggregate judgments, opinions, or predictions from a large group of individuals for improved decision making. However, existing literature mostly focuses on eliciting the wisdom of crowds in an offline context—without tapping into the vast amount of data available on online crowdsourcing platforms. To extract WoC from participants on online platforms, there exist at least three challenges, including social influence, suboptimal aggregation strategies, and data sparsity. This dissertation aims to answer the research question of how to effectively extract WoC from crowdsourcing platforms for the purpose of making better decisions. In the first study, I designed a new opinions aggregation method, Social Crowd IQ (SCIQ), using a time-based decay function to eliminate the impact of social influence on crowd performance. In the second study, I proposed a statistical learning method, CrowdBoosting, instead of a heuristic-based method, to improve the quality of crowd wisdom. In the third study, I designed a new method, Collective Persuasibility, to solve the challenge of data sparsity in a crowdfunding platform by inferring the backers' preferences and persuasibility. My work shows that people can obtain business benefits from crowd wisdom, and it provides several effective methods to extract wisdom from online crowdsourcing platforms, such as StockTwits, Good Judgment Open, and Kickstarter. / Doctor of Philosophy / Since Web 2.0 and mobile technologies have inspired increasing numbers of people to contribute and interact online, crowdsourcing provides a great opportunity for the businesses to tap into a large group of online users who possess varied capabilities, creativity, and knowledge levels. Howe (2006) first defined crowdsourcing as a method for obtaining necessary ideas, information, or services by asking for contributions from a large group of individuals, especially participants in online communities. Many online platforms have been developed to support various crowdsourcing tasks, including crowdfunding (e.g., Kickstarter and Indiegogo), crowd prediction (e.g., StockTwits, Good Judgment Open, and Estimize), crowd creativity (e.g., Wikipedia), and crowdsolving (e.g., Dell IdeaStorm). The explosive data generated by those platforms give us a good opportunity for business benefits. Specifically, guided by the Wisdom of Crowds (WoC) theory, we can aggregate multiple opinions from a crowd of individuals for improving decision making. In this dissertation, I apply WoC to three crowdsourcing tasks, stock return prediction, event outcome forecast, and crowdfunding project success prediction. Our study shows the effectiveness of WoC and makes both theoretical and practical contributions to the literature of WoC.

Dans cette thèse, nous nous sommes intéressés à l'impact de la quantité et de la qualité de l'information échangée entre individus d'un groupe sur leurs performances collectives dans deux types de tâches bien spécifiques. Dans une première série d'expériences, les sujets devaient estimer des quantités séquentiellement, et pouvaient réviser leurs estimations après avoir reçu comme information sociale l'estimation moyenne d'autres sujets. Nous contrôlions cette information sociale à l'aide de participants virtuels (dont nous contrôlions le nombre) donnant une information (dont nous contrôlions la valeur), à l'insu des sujets. Nous avons montré que lorsque les sujets ont peu de connaissance préalable sur une quantité à estimer, (les logarithmes de) leurs estimations suivent une distribution de Laplace. La médiane étant un bon estimateur du centre d'une distribution de Laplace, nous avons défini la performance collective comme la proximité de la médiane (du logarithme) des estimations à la vraie valeur. Nous avons trouvé qu'après influence sociale, et lorsque les agents virtuels fournissent une information correcte, la performance collective augmente avec la quantité d'information fournie (fraction d'agents virtuels). Nous avons aussi analysé la sensibilité à l'influence sociale des sujets, et trouvé que celle-ci augmente avec la distance entre l'estimation personnelle et l'information sociale. Ces analyses ont permis de définir 5 traits de comportement : garder son opinion, adopter celle des autres, faire un compromis, amplifier l'information sociale ou au contraire la contredire. Nos résultats montrent que les sujets qui adoptent l'opinion des autres sont ceux qui améliorent le mieux leur performance, car ils sont capables de bénéficier de l'information apportée par les agents virtuels. Nous avons ensuite utilisé ces analyses pour construire et calibrer un modèle d'estimation collective, qui reproduit quantitativement les résultats expérimentaux et prédit qu'une quantité limitée d'information incorrecte peut contrebalancer un biais cognitif des sujets consistant à sous-estimer les quantités, et ainsi améliorer la performance collective. D'autres expériences ont permis de valider cette prédiction. Dans une seconde série d'expériences, des groupes de 22 piétons devaient se séparer en clusters de la même "couleur", sans indice visuel (les couleurs étaient inconnues), après une courte période de marche aléatoire. Pour les aider à accomplir leur tâche, nous avons utilisé un système de filtrage de l'information disponible (analogue à un dispositif sensoriel tel que la rétine), prenant en entrée l'ensemble des positions et couleurs des individus, et retournant un signal sonore aux sujets (émit par des tags attachés à leurs épaules) lorsque la majorité de leurs k plus proches voisins était de l'autre couleur que la leur. La règle consistait à s'arrêter de marcher lorsque le signal stoppait. / In this thesis, we were interested in the impact of the quantity and quality of information ex- changed between individuals in a group on their collective performance in two very specific types of tasks. In a first series of experiments, subjects had to estimate quantities sequentially, and could revise their estimates after receiving the average estimate of other subjects as social information. We controlled this social information through virtual participants (which number we controlled) giving information (which value we controlled), unknowingly to the subjects. We showed that when subjects have little prior knowledge about a quantity to estimate, (the loga- rithms of) their estimates follow a Laplace distribution. Since the median is a good estimator of the center of a Laplace distribution, we defined collective performance as the proximity of the median (log) estimate to the true value. We found that after social influence, and when the information provided by the virtual agents is correct, the collective performance increases with the amount of information provided (fraction of virtual agents). We also analysed subjects' sensitivity to social influence, and found that it increases with the distance between personal estimate and social information. These analyses made it possible to define five behavioral traits: to keep one's opinion, to adopt that of others, to compromise, to amplify social information or to contradict it. Our results showed that the subjects who adopt the opinion of others are the ones who best improve their performance because they are able to benefit from the infor- mation provided by the virtual agents. We then used these analyses to construct and calibrate a model of collective estimation, which quantitatively reproduced the experimental results and predicted that a limited amount of incorrect information can counterbalance a cognitive bias that makes subjects underestimate quantities, and thus improve collective performance. Further experiments have validated this prediction. In a second series of experiments, groups of 22 pedestrians had to segregate into clusters of the same "color", without visual cue (the colors were unknown), after a short period of random walk. To help them accomplish their task, we used an information filtering system (analogous to a sensory device such as the retina), taking all the positions and colors of individuals in input, and returning an acoustic signal to the subjects (emitted by tags attached to their shoulders) when the majority of their k nearest neighbors was of a different color from theirs.

Many animal groups benefit from making decisions collectively. For example, colonies of many ant species are able to select the best possible nest to move into without every ant needing to visit each available nest site. Similarly, honey bee colonies can focus their foraging resources on the best possible food sources in their environment by sharing information with each other. In the same way, groups of human individuals are often able to make better decisions together than each individual group member can on his or her own. This phenomenon is known as "collective intelligence", or "wisdom of crowds." What unites all these examples is the fact that there is no centralized organization dictating how animal groups make their decisions. Instead, these successful decisions emerge from interactions and information transfer between individual members of the group and between individuals and their environment. In this thesis, I apply mathematical modeling techniques in order to better understand how groups of social animals make important decisions in situations where no single individual has complete information. This thesis consists of five papers, in which I collaborate with biologists and sociologists to simulate the results of their experiments on group decision-making in animals. The goal of the modeling process is to better understand the underlying mechanisms of interaction that allow animal groups to make accurate decisions that are vital to their survival. Mathematical models also allow us to make predictions about collective decisions made by animal groups that have not yet been studied experimentally or that cannot be easily studied. The combination of mathematical modeling and experimentation gives us a better insight into the benefits and drawbacks of collective decision making, and into the variety of mechanisms that are responsible for collective intelligence in animals. The models that I use in the thesis include differential equation models, agent-based models, stochastic models, and spatially explicit models. The biological systems studied included foraging honey bee colonies, house-hunting ants, and humans answering trivia questions.

Crowdsourcing can today be found in practically any industry, but the extent to which it is used differ widely. A report from last year, published by the crowdsourcing platform eYeka (eYeka 2015b), shows that the fashion industry is among the industries using crowdsourcing the least. Brands that are more inclined to using crowdsourcing are those working with fast moving consumer goods (FMCG). That includes many brands from the food &amp; beverage industry such as Coca-Cola, Pepsi, Danone etc. This study builds on this knowledge to explore the use of crowdsourcing in the food &amp; beverage industry to find out if that experience can be used in the fashion industry where it is not widely used today. To identify different approaches of crowdsourcing used in the food &amp; beverage industry 78 crowdsourcing campaigns from 9 brand during a two years period (2014-2015) was analysed. The analysis resulted in the identification of 3 main approaches: crowdsourcing as ideation, customer engagement and crowdsourcing for creation/production. More importantly this study comes to the conclusion that the way crowdsourcing is used by the brands in the food &amp; beverage industry is not formed in such a way that it is exclusive to the industry in question. There are no immediate boundaries for the fashion industry to adopt the same way of working. If the fashion industry would like to follow the trend in food &amp; beverage industry they should focus more on the creative ideation side of crowdsourcing and less on the creation, and most of all on marketing. However, there are also other areas they can learn from like Business Development.

Wie Menschen von anderen lernen und wann es adaptiv-rational ist sich auf soziales Lernen zu verlassen sind wichtige Fragen in vielen Disziplinen einschließlich der Psychologie, der Biologie, der Anthropologie und den Wirtschaftswissenschaften. Trotz der geteilten Interessen dieser Disziplinen sind viele der vorhandenen Resultate voneinander isoliert und oft nicht vergleichbar, teilweise weil es der Forschung zum sozialen Lernen immer noch eines theoretischen Rahmens fehlt, welcher die gewonnen Erkenntnisse vergleichbar machen würde sowie erklären würde warum unterschiedliche Strategien in Abhängigkeit vom sozialen Kontext erfolgreich sind oder nicht. In meiner Arbeit schlage ich einen solchen theoretischen Rahmen vor, welcher sich auf der Forschung zur ökologischen Rationalität gründet. Ich benutze den theoretischen Rahmen der ökologischen Rationalität sozialen Lernens, um drei Fragen zu beantworten: i) Wie können soziale Lernstrategien als kognitiv plausible Strategien modelliert werden, die auf drei einfachen Building Blocks beruhen (Such-, Stopp- und Entscheidungsregeln), ii) was sind die wichtigsten Faktoren von sozialen Umwelten und Problemumwelten, in denen soziales Lernen stattfindet und iii) wie interagieren soziale Lernstrategien, die auf unterschiedlichen Building Blocks beruhen, mit der Struktur von Umwelten, um unterschiedliche Erfolgsniveaus zu erreichen. Indem ich diese drei Fragen adressiere, erarbeite ich die Bedingungen unter denen unterschiedlichen Strategien adaptiv-rational sind und erkläre wie unterschiedlichen Strategien in bestimmten Umwelten erfolgreich sind. Jedes der Kapitel behandelt eine wichtige alltägliche soziale Lernsituation, identifiziert die Schlüsselcharakteristiken der Situation und demonstriert wie die Building Blocks des sozialen Lernens mit diesen Umweltstrukturen interagieren, um unterschiedliche Erfolgsniveaus zu erreichen. / How people learn from others and when it is adaptive to rely on social learning have been major questions in several disciplines including psychology, biology, anthropology and economics. Despite the shared interest of these diverse fields, many of the results remain isolated and are often incomparable, in part because the study of social learning still lacks a general theoretical framework that would make results comparable or explain why different strategies perform well in different contexts. In this thesis I propose such a framework that is grounded in the study of ecological rationality. I use this frame- work to explore three primary questions: i) how can social learning strategies be modeled as cognitively plausible strategies composed of simple building blocks (search, stopping and decision rules), ii) what are key characteristics of social and task environments in which social learning takes place, and iii) how do social learning strategies composed of different building blocks interact with the structure of the environment to produce different levels of success. Through addressing these three questions I map out the conditions under which different strategies are adaptive and explain how the building blocks of different strategies contribute to their performance in certain environments. The thesis focuses on three representative classes of social learning strategies, namely, frequency-dependent, payoff-biased, and unbiased copying. Different chapters focus on important everyday social learning settings, identify key environmental characteristics defining the setting and demonstrate how the building blocks of social learning strategies interact with these environmental structures to produce different outcomes.

No description available.

The diploma thesis "Estimation of the election results by respondents and its usage" discuss the methods of predicting the election results based on election expectations of individuals and shows how people form their expectations and how exact these expectations are. First short summary of existing methods of election results predictions and its limitations is presented - it also deals with pre-election surveys and its ambitions to predict the election results. The rest of the thesis focuses only on the prediction of the election results based on election expectations of individuals: prediction markets and aggregated estimations of respondents. Concept Wisdom of Crowds, from which both approaches originate, is presented together with concrete examples of application of the predictions based on opinions of prediction markets participants and respondents of the pre-election surveys. Results of the foreign studies confirm that the prediction markets predict the election results very well and with higher accuracy than the pre-election surveys. Current studies also positively evaluate the estimation of the election results done by respondents. Respondents are usually able to predict the election results, even several weeks before the elections. Last part of the thesis is based on my own quantitative...